Re: [nc-whois] gTLD Thinking on the Way Ahead

[Thomas Roessler <roessler@does-not-exist.org>]

On 2002-02-27 09:54:06 -0000, Karen Elizaga wrote:

>We'd like to make some general suggestions about the best path 
>ahead, in light of our conference call later today, as well as 
>specific comments on the draft.

Thank you for your very timely comments.

>With respect to the final whois report, we feel strongly that it 
>should reflect careful, thoughtful analysis of all the data.  This 
>is important if our recommendations are to gain support from the 
>larger community. For that reason, we continue to believe, as the 
>TF decided a few months ago, that we owe it to the people who took 
>the time to respond to the survey to read what they wrote. We 
>realize this is a time-consuming effort, but we can divide up the 
>work.  (We are not certain if others are suggesting that we 
>revisit this issue, but in case so we wanted to restate our view.)

With all due respect, consensus process should not be used as an 
excuse to withhold obvious results from publication.  Also, just 
like we may owe respondents to individually look at all the 
free-text responses, we owe them that whatever can be derived from 
the report be published in a timely manner.

Also, if you are really so worried about bottom-up process, why 
don't we just publish the data, so members of the community can come 
to their own conclusions?

>With respect to the Ghana report, we think it is important that it 
>convey a good overview of the statistical data that we have. 

Indeed.  This should, in particular, imply that we don't withhold 
part of the insight we have gained.  (But I don't believe you wanted 
to suggest that...)

>We envision (1) a summary of where we are in the process of 
>preparing a final report, (2) an overview of the statistical 
>breakdown, including observations about different audiences 
>(including the many instances where they do not differ) - Note 
>this can be done relatively easily from the 22 page breakdown 
>ICANN sent us - we suggest dividing the questions up by 
>constituency for efficiency;

Abel has already produced a large excel sheet with all these data, 
which can readily be used for this task.  It's really just about 
copying the numbers, and generating charts from these.  A good 
secretary can prepare the final charts from this in half a day. 
(I'm not sure Abel posted his numbers to this list, though.)

>With respect to Thomas' draft, we think it seems more appropriate 
>for the final report, after we have been through all the data, 
>than for the interim report. As mentioned above, we envision a 
>short description of each issue where have analyzed free text, 
>e.g., bulk access and the others.  But at this stage it would be 
>premature to draw "conclusions" without looking in detail at all 
>the questions and all the surveys.  

I strongly disagree.  

But maybe you can point out how an analysis of the free-form 
responses and the overall sample should lead to conclusions 
different from the ones I've been presenting in my draft?  I'd 
welcome any evidence for this, which should surely be present in the 
"statistical 303" set we have.

Lacking such evidence, I don't see any reason why we should withhold 
our results.  After all, your constituency is interested in getting 
results out of this task force, isn't it?

>that (1) we should clarify whether we are referring to individuals 
>only (vs commercial entities); 

I think that my draft is extremely clear on where I'm talking about 
all respondents, or about respondents of certain categories.  In 
particular, where I mentioned consensus over categories, this always 
was consensus which included individual participants _and_ 
commercial entities.

>(2) the statistics on p4 appear do not reconcile with the ICANN 
>data we received; 

I just double-checked the numbers I've used.  They match the ones 
 from the Whois Survey Results distributed.

If you are referring to the percentages (which do differ), you may 
notice that ICANN has included the "no response" answers.  The 
percentages I've been using are only referring to those respondents 
who actually answered the question.  The results from the set of 303 
questionnaires were generated by automatic counting (using standard 
Unix tools such as grep, sort, and uniq).  You are, of course, 
welcome to double-check these numbers if you doubt them.  Happy 
counting.

Note that the other tables were generated by just cutting & pasting 
(and properly formatting) the data from ICANN's tabulations, so 
there should be no notable differences, with the exception of giving 
percentages of those who actually responded to a particular 
question, as opposed to giving percentages of all participants in 
the survey.

>(3) it may not be possible to simply add the statistical data on 
>p5 to produce the 3rd chart; 

I suppose you are talking about the 3rd _table_.  Since the answers 
were mutually exclusive, cardinalities really just add up: Denote 
unions of sets with +, and intersections with *.  Consider two sets 
A, B, and denote by #X the number of elements in a set X. Then,

	#(A + B) = #A + #B - #(A*B).

However, since A*B is empty in our case, we just end up with plain 
addition.  (That's primary school maths, folks!)

Also note that I have carefully avoided to compare or add up 
absolute numbers from different rows, or to add up percentages from 
different rows.

Thus, I don't believe that I've done any comparison of apples and 
oranges in the draft.

>and (4) the contradiction between questions 16 and 17 needs to be 
>explored more fully.

I'm looking forward to discuss this during today's conference call. 
I do believe that the argument I've been giving in my draft is 
sound, though.  We'll see what the rest of the task force thinks 
about this question.

>In addition, both the interim and the final reports need to be 
>honest about what the survey does do and doesn't do, and its 
>shortcomings, now that those of us who helped write have the 
>benefit of hindsight

Indeed.  I'm already preparing a statistics part which I hope to 
prepare in time for today's conference call.  I'll have to polish 
some of my statistical knowledge, though.

Thanks,
-- 
Thomas Roessler                        http://log.does-not-exist.org/